Predicting the evolution of Escherichia coli by a data-driven approach.

Area of Science:

• Evolutionary Biology
• Genomics
• Microbial Evolution

Background:

• Predicting evolutionary trajectories remains a significant challenge in biology.
• Understanding mutation patterns under diverse environmental conditions is crucial for predicting adaptation.

Purpose of the Study:

• To determine if evolution can be predicted from historical mutation data.
• To build predictive models for gene-level evolution in Escherichia coli.

Main Methods:

• Compiled a compendium of over 15,000 mutation events from Escherichia coli across 178 environments.
• Analyzed mutation hotspots, co-occurrence, and shared mutations across replicates.
• Trained ensemble predictors on the mutation compendium and validated them on forward evolution experiments.

Main Results:

• Mutation compendium analysis revealed patterns of mutation hotspots and co-occurrence.
• Pairwise mutation overlap ratio remained consistent irrespective of replicate number.
• Predictive models achieved 49.2% precision and 34.5% recall in identifying mutation targets.

Conclusions:

• Integrated datasets of mutation events can be used to develop predictive models of evolution.
• Predictive models offer insights into evolutionary processes in controlled environments.
• This approach advances the predictability of microbial evolution at the gene level.